Affine Global Motion Estimation and Hardware Implementation Mong

Affine Global Motion Estimation and Hardware Implementation Mong Chit Wong (Maggie) and Prof. Truong Nguyen Video Processing Group, UCSD Hardware Implementation (coarse estimation) Network Channel GME coding Done Address_read Address_write Enable_m 1 Go Decimation Module Enable_m 2 FFT Module Done Input I, R Output Magnitude Module Reset_n Go Log polar Module Reset_n Transformed Image Refined Parameters Phase Correlation Estimate of B Affine compensation Old spec Done New spec Go Cross Power Spectrum Module Estimation of A Coarse Affine Compensation - Normalization approximation - Phase plane correlation optimization Log-polar Mapping Refinement of Parameters in Spatial domain - Log base 2 approximation Log 2(13) = 3. 7 Log 2(1101) = 3 + ½ + 1/8 = 3. 625 ~ log 2(13) FFT Module Translation Invariance Decimation Coarse estimation In Frequency domain Go done In xy - Magnitude approximation Out log polar Coarse Estimation in Frequency domain Reference Image Optimization Tricks: Different tricks are used in the implementation to reduce mathematical complexity. Output frame Enable_m Reference Transformed Image Reset_n State 1 _nstate 2 Enable_m 3 Out address Proposed Method The hardware is being implemented in verilog. Each of the modules are simulated in Modelsim 5. 8 and Matlab to verify the function correctness and bit length, and eventually synthesized by Cadence Build gate. Input frame Control Visual communication has evolved rapidly due to the hardware improvement and compression technology. Videos are highly compressible because of the similarity in their neighboring frames. Our novel GME algorithm is able to track the background information using merely a 6 parameter affine model per frame. Compared to the conventional motion estimation used in MPEG 2 which uses block-based matching and uses more than hundreds of motion vectors per frame, the new GME algorithm uses only one motion vector per frame. And thus achieved higher coding efficiency and better video results. We carried out hardware implementation on the coarse estimation step of the GME algorithm during the summer research period. This is on-going research, and further implementation and optimization on the rest of the steps are expected in the future. Memory Abstract Estimate of A Reset_n Reserved spec Output wave from Modelsim (decimator) Inverse FFT Module Simulation Results Synthesized Result (By Cadence Buildgates) Go Decimation Module – 74000 gate counts FFT Magnitude – 7800 gate counts Log – polar – 200000 gate counts Reset_n Speed – 100 MHz done correlation index Peak Location Module Initial error image Dataflow from Modelsim (decimator) After coarse estimation After refinement Coarse estimated motion vectors